Oct 21, 2014

We have all seen weather forecasts where the chance of precipitation is given as a percentage. For example, tomorrow there is an 80% chance of rain showers during the day. For many in the weather community it is well-known exactly what this percentage represents, however, in the public there can sometimes be some confusion and misunderstandings. So let’s break down some misconceptions people may have about probability of precipitation:

**Misconception: The percentage given represents the area over which it will rain.**

Sometimes people take the percentage and misapply it as the percentage of an area that will see rain. It is incorrectly assumed that the an 80% chance of rain means that 80% of a forecast area (ex. the town of Poughkeepsie, NY) will see rain on the given day or night. Meteorologists are able to predict the coverage of precipitation within their forecast area, however, this is not how probability of precipitation is determined.

**Misconception: The percentage given represents the duration of the precipitation event.**

This is another common, and easily understood myth surrounding precipitation probability. Often people will see that there is a 30% chance of rain during the day and believe that it means that it will only rain for 30% of the day and the rest will be dry. While this is a logical conclusion, it is not correct. Instead, meteorologists will address the duration of a precipitation event in a forecast discussion or in the summary. Ex. 30% chance of showers, mainly before 10am.

**Reality: The percentage given represents the chance (probability) of any amount of precipitation.**

The probability percentage is applied to the total forecast area (city, county, region, etc.) and is valid for the entire forecast period (generally daytime 7am – 7am and nighttime 7pm – 7am). So a forecast with a 40% chance of rain overnight means that the probability that any area within the forecast zone, during the overnight hours observes a measurable amount of precipitation, is 40%.

Perhaps one of my biggest pet peeves in the meteorological community is forecasting 50% chance of precipitation. Essentially what this is telling your audience is that you, as the meteorologist, have no idea whether it will or will not rain. By forecasting a 50% chance, the meteorologist is absolving themselves from any sort of responsibility or accountability in regards to the accuracy of their forecast. There is no skill involved in forecasting a 50% chance of precipitation, because what it really comes down to is a toss of a coin.

With all the resources at our disposal these days, such as computer models, satellite imagery, real-time surface observations, 12-hr radiosonde launches and high resolution doppler radar, we should be able to provide the public with a forecast that is better than a coin toss. Truthfully, it is an embarrassment to our industry when meteorologists choose to consistently forecast 50%. Whether it precipitates or not, the meteorologist could claim that he or she was correct, because they did not take a stance in either direction. In my opinion, any meteorologist that forecasts 50% is wrong, no matter what happens.

I will note that there are occasions when predicting 50/50 is a legitimate option, but such occurrences should be extremely rare and the meteorologist in question should provide plenty of reasoning behind their decision to ride the fence.

When it comes to forecasting precipitation, a meteorologist should always try to fall on either side of the fence. After reviewing all the available data, does it seem more or less likely than not to precipitate? If you are close to the 50% range, try to fall on either side of it. Most of us meteorologists have spent a great deal of time and effort learning the skills of weather forecasting and I believe it is our responsibility to be better than a coin toss.

Here at Austin’s Atmosphere I regularly verify and score my own forecasts. Every time I issue a forecast I enter the relevant variables (min/max temp, probability of precip, total liquid equivalent accumulation, and max sustained wind speed) into a specially-designed Excel spreadsheet. A series of macros that I wrote are executed and pull actual observed conditions for the nearest official weather observation station (ex. KPOU in Poughkeepsie, NY). Those actual observations are then passed through a series of filters and equations and then the difference between my forecast values and the actual values is calculated and given a corresponding point value. Each forecast period is scored and the total error points are summed up. The lower the score, the better my forecast.

When it comes to scoring probability of precipitation, the actual observation is either “yes” or “no”, it either did rain or it didn’t. In the forecasting section I would enter a 3 for 30% and a 10 for 100%. In the actual section, if .01″ of rain or more is observed then the actual value is 10, if no precipitation is recorded at the station then the actual value is 0. There is no in-between values when it comes to actual, it either rained or it didn’t. My scoring approach is as follows:

|forecast PoP – actual|^2 = total PoP error points

An example of this: I forecast 40% chance of rain and actual is 0.00″ the score would be calculated like this:

|4 – 0|^2 = 4^2 = 16 total PoP error points

Another example: I forecast 90% chance of rain and actual is greater than 0.00″ the score would be calculated like this:

|9 – 10|^2 = 1^2 = 1 total PoP error point

As you can see this method of scoring PoP (probability of precipitation) penalizes the forecaster more harshly the farther from actual their forecast value was. So if you said there was a 0% chance of rain, and it did rain, even .01″, then you would be penalized with 100 error points! And rightfully so.

In the end, I would simply like to see more meteorologists get away from using the 50% probability in their forecasts. 50% gives no value or credibility to their work and speaks badly of their forecasting skill. I would rather forecast 40% and be wrong then forecast 50% and be wrong either way. Demonstrate your forecasting skill, don’t ride the 50/50 fence. Even if you get burned now and then you will hopefully gain valuable experience and increase your skills with time.

You must log in to post a comment.